Skip to contentSkip to conversation

selinux: Add helper functions to get and set checkreqprot

PATCH Superseded Lakshmi Ramasubramanian 3 participants 7 messages Sep 9, 2020 – Sep 11, 2020 SELinux security module

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 selinux: Add helper functions to get and set checkreqprot

checkreqprot data member in selinux_state struct is accessed directly by SELinux functions to get and set. This could cause unexpected read or write access to this data member due to compiler optimizations and/or compiler's reordering of access to this field.

Add helper functions to get and set checkreqprot data member in selinux_state struct. These helper functions use READ_ONCE and WRITE_ONCE macros to ensure atomic read or write of memory for this data member.

---
 security/selinux/hooks.c            |  6 +++---
 security/selinux/include/security.h | 10 ++++++++++
 security/selinux/selinuxfs.c        |  5 +++--
 3 files changed, 16 insertions(+), 5 deletions(-)
security/selinux/hooks.c +3 -3
index 6210e98219a5..25a36a3e6bed 100644
@@ -3718,7 +3718,7 @@ static int selinux_mmap_file(struct file *file, unsigned long reqprot,
 			return rc;
 	}
 
-	if (selinux_state.checkreqprot)
+	if (checkreqprot_enabled(&selinux_state))
 		prot = reqprot;
 
 	return file_map_prot_check(file, prot,
@@ -3732,7 +3732,7 @@ static int selinux_file_mprotect(struct vm_area_struct *vma,
 	const struct cred *cred = current_cred();
 	u32 sid = cred_sid(cred);
 
-	if (selinux_state.checkreqprot)
+	if (checkreqprot_enabled(&selinux_state))
 		prot = reqprot;
 
 	if (default_noexec &&
@@ -7234,7 +7234,7 @@ static __init int selinux_init(void)
 
 	memset(&selinux_state, 0, sizeof(selinux_state));
 	enforcing_set(&selinux_state, selinux_enforcing_boot);
-	selinux_state.checkreqprot = selinux_checkreqprot_boot;
+	checkreqprot_set(&selinux_state, selinux_checkreqprot_boot);
 	selinux_avc_init(&selinux_state.avc);
 	mutex_init(&selinux_state.status_lock);
 	mutex_init(&selinux_state.policy_mutex);
security/selinux/include/security.h +10 -0
index cbdd3c7aff8b..cc29177c8858 100644
@@ -143,6 +143,16 @@ static inline void enforcing_set(struct selinux_state *state, bool value)
 }
 #endif
 
+static inline bool checkreqprot_enabled(const struct selinux_state *state)
+{
+	return READ_ONCE(state->checkreqprot);
+}
+
+static inline void checkreqprot_set(struct selinux_state *state, bool value)
+{
+	WRITE_ONCE(state->checkreqprot, value);
+}
+
 #ifdef CONFIG_SECURITY_SELINUX_DISABLE
 static inline bool selinux_disabled(struct selinux_state *state)
 {
security/selinux/selinuxfs.c +3 -2
index 45e9efa9bf5b..540bfa078877 100644
@@ -717,7 +717,8 @@ static ssize_t sel_read_checkreqprot(struct file *filp, char __user *buf,
 	char tmpbuf[TMPBUFLEN];
 	ssize_t length;
 
-	length = scnprintf(tmpbuf, TMPBUFLEN, "%u", fsi->state->checkreqprot);
+	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
+			   checkreqprot_enabled(fsi->state));
 	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
 }
 
@@ -759,7 +760,7 @@ static ssize_t sel_write_checkreqprot(struct file *file, const char __user *buf,
 			     comm, current->pid);
 	}
 
-	fsi->state->checkreqprot = new_value ? 1 : 0;
+	checkreqprot_set(fsi->state, (new_value ? 1 : 0));
 	length = count;
 out:
 	kfree(page);
Signature · 1 line
2.28.0
maintainer Acked-by on v2 1/1
#1 lore raw
Quoted 6 lines · Lakshmi Ramasubramanian

checkreqprot data member in selinux_state struct is accessed directly by SELinux functions to get and set. This could cause unexpected read or write access to this data member due to compiler optimizations and/or compiler's reordering of access to this field.

Quoted 7 lines

Add helper functions to get and set checkreqprot data member in selinux_state struct. These helper functions use READ_ONCE and WRITE_ONCE macros to ensure atomic read or write of memory for this data member.

maintainer
#2 lore raw
In reply to Lakshmi Ramasubramanian
Quoted 18 lines · Lakshmi Ramasubramanian

checkreqprot data member in selinux_state struct is accessed directly by SELinux functions to get and set. This could cause unexpected read or write access to this data member due to compiler optimizations and/or compiler's reordering of access to this field.

Add helper functions to get and set checkreqprot data member in selinux_state struct. These helper functions use READ_ONCE and WRITE_ONCE macros to ensure atomic read or write of memory for this data member.

---
 security/selinux/hooks.c            |  6 +++---
 security/selinux/include/security.h | 10 ++++++++++
 security/selinux/selinuxfs.c        |  5 +++--
 3 files changed, 16 insertions(+), 5 deletions(-)

...

Quoted 7 lines
security/selinux/include/security.h +0 -0
index cbdd3c7aff8b..cc29177c8858 100644
@@ -143,6 +143,16 @@ static inline void enforcing_set(struct selinux_state *state, bool value)
 }
 #endif

security/selinux/include/security.h static inline void enforcing_set(struct selinux_state *state, bool value) · line 154

Quoted 10 lines
security/selinux/include/security.h +9 -0
 
+static inline bool checkreqprot_enabled(const struct selinux_state *state)
+{
+       return READ_ONCE(state->checkreqprot);
+}
+
+static inline void checkreqprot_set(struct selinux_state *state, bool value)
+{
+       WRITE_ONCE(state->checkreqprot, value);
+}

This is a nitpick, and I recognize that Stephen already suggested the use of "*_set()" and "*_enabled()" for names, but if we are going to name the setter "*_set()" let's also name the getter "*_get()".

Other than that, it looks fine to me.

Signature · 2 lines
paul moore
www.paul-moore.com
maintainer
#3 lore raw
Quoted 32 lines · Paul Moore
Quoted 18 lines · Lakshmi Ramasubramanian

checkreqprot data member in selinux_state struct is accessed directly by SELinux functions to get and set. This could cause unexpected read or write access to this data member due to compiler optimizations and/or compiler's reordering of access to this field.

Add helper functions to get and set checkreqprot data member in selinux_state struct. These helper functions use READ_ONCE and WRITE_ONCE macros to ensure atomic read or write of memory for this data member.

---
 security/selinux/hooks.c            |  6 +++---
 security/selinux/include/security.h | 10 ++++++++++
 security/selinux/selinuxfs.c        |  5 +++--
 3 files changed, 16 insertions(+), 5 deletions(-)

...

Quoted 8 lines
security/selinux/include/security.h +0 -0
index cbdd3c7aff8b..cc29177c8858 100644
@@ -143,6 +143,16 @@ static inline void enforcing_set(struct selinux_state *state, bool value)
 }
 #endif
 
Quoted 13 lines

security/selinux/include/security.h static inline void enforcing_set(struct selinux_state *state, bool value) · line 154

Quoted 9 lines
security/selinux/include/security.h +9 -0
+static inline bool checkreqprot_enabled(const struct selinux_state *state)
+{
+       return READ_ONCE(state->checkreqprot);
+}
+
+static inline void checkreqprot_set(struct selinux_state *state, bool value)
+{
+       WRITE_ONCE(state->checkreqprot, value);
+}

This is a nitpick, and I recognize that Stephen already suggested the use of "*_set()" and "*_enabled()" for names, but if we are going to name the setter "*_set()" let's also name the getter "*_get()".

I just suggested that we be consistent with the existing naming for enforcing_*(), which I thought came from you?

maintainer
#4 lore raw
Quoted 46 lines · Stephen Smalley
Quoted 31 lines · Paul Moore
Quoted 18 lines · Lakshmi Ramasubramanian

checkreqprot data member in selinux_state struct is accessed directly by SELinux functions to get and set. This could cause unexpected read or write access to this data member due to compiler optimizations and/or compiler's reordering of access to this field.

Add helper functions to get and set checkreqprot data member in selinux_state struct. These helper functions use READ_ONCE and WRITE_ONCE macros to ensure atomic read or write of memory for this data member.

---
 security/selinux/hooks.c            |  6 +++---
 security/selinux/include/security.h | 10 ++++++++++
 security/selinux/selinuxfs.c        |  5 +++--
 3 files changed, 16 insertions(+), 5 deletions(-)

...

Quoted 8 lines
security/selinux/include/security.h +0 -0
index cbdd3c7aff8b..cc29177c8858 100644
@@ -143,6 +143,16 @@ static inline void enforcing_set(struct selinux_state *state, bool value)
 }
 #endif
 
Quoted 13 lines

security/selinux/include/security.h static inline void enforcing_set(struct selinux_state *state, bool value) · line 154

Quoted 9 lines
security/selinux/include/security.h +9 -0
+static inline bool checkreqprot_enabled(const struct selinux_state *state)
+{
+       return READ_ONCE(state->checkreqprot);
+}
+
+static inline void checkreqprot_set(struct selinux_state *state, bool value)
+{
+       WRITE_ONCE(state->checkreqprot, value);
+}

This is a nitpick, and I recognize that Stephen already suggested the use of "*_set()" and "*_enabled()" for names, but if we are going to name the setter "*_set()" let's also name the getter "*_get()".

Quoted 2 lines

I just suggested that we be consistent with the existing naming for enforcing_*(), which I thought came from you?

I vaguely remember having a discussion around the naming back then, but I can't recall who preferred what and when. I'm sure it's all in the archives for anyone who cares to look.

Regardless, I think _set() and _get() make the most sense here. I can also see an argument being made for "_enabled()" on some bigger flags. My personal opinion is that the SELinux kernel code, and kernel code in general, is a bit of a mess when it comes to naming consistency. Considering the age, size, and number of contributors the current state really shouldn't be too surprising (and honestly it isn't that bad considering everything). Perhaps someday we can look at that, but that is so far down the priority list it isn't worth discussing; I'd tackle the coding style inconsistencies before this.

Signature · 2 lines
paul moore
www.paul-moore.com
#5 lore raw
In reply to Paul Moore
Quoted 15 lines · Paul Moore

security/selinux/include/security.h static inline void enforcing_set(struct selinux_state *state, bool value) · line 154

Quoted 10 lines
security/selinux/include/security.h +9 -0
+static inline bool checkreqprot_enabled(const struct selinux_state *state)
+{
+       return READ_ONCE(state->checkreqprot);
+}
+
+static inline void checkreqprot_set(struct selinux_state *state, bool value)
+{
+       WRITE_ONCE(state->checkreqprot, value);
+}

This is a nitpick, and I recognize that Stephen already suggested the use of "*_set()" and "*_enabled()" for names, but if we are going to name the setter "*_set()" let's also name the getter "*_get()".

Quoted 2 lines

Other than that, it looks fine to me.

Sure - I can do that.

Are you expecting something like below (for checkreqprot and enforcing)?

s/checkreqprot_enabled/checkreqprot_get/

s/enforcing_enabled/enforcing_get/

-lakshmi

maintainer
#6 lore raw
Quoted 19 lines · Lakshmi Ramasubramanian
Quoted 15 lines · Paul Moore

security/selinux/include/security.h static inline void enforcing_set(struct selinux_state *state, bool value) · line 154

Quoted 10 lines
security/selinux/include/security.h +9 -0
+static inline bool checkreqprot_enabled(const struct selinux_state *state)
+{
+       return READ_ONCE(state->checkreqprot);
+}
+
+static inline void checkreqprot_set(struct selinux_state *state, bool value)
+{
+       WRITE_ONCE(state->checkreqprot, value);
+}

This is a nitpick, and I recognize that Stephen already suggested the use of "*_set()" and "*_enabled()" for names, but if we are going to name the setter "*_set()" let's also name the getter "*_get()".

Quoted 2 lines

Other than that, it looks fine to me.

Quoted 7 lines

Sure - I can do that.

Are you expecting something like below (for checkreqprot and enforcing)?

s/checkreqprot_enabled/checkreqprot_get/

s/enforcing_enabled/enforcing_get/

Sorry for the confusion, I should have been more clear. I was thinking that the names "checkreqprot_set()" and "checkreqprot_get()" would be nice.

Signature · 2 lines
paul moore
www.paul-moore.com